AVEVA PI WORLD Distillation Column Convergence in AVEVA ...

Distillation Column Convergence in AVEVA™ PRO/II™ Simulation

AVEVA PI WORLD

© 2021 AVEVA Group plc and its subsidiaries. All rights reserved.

E-4-S/L

Mike Donahue

10/21/2021


• Convergence

• Damping

• Estimates

• Initial Specifications

• Algorithms

• Errors

• Strategy

Agenda

Convergence



BackgroundEquilibrium Stages


ConvergenceEquations

• Mass and energy must be conserved

• The vapor and liquid leaving a stage must be in equilibrium

• The vapor and liquid mole fractions must sum to 1

M Mass

E Equilibrium

S Summation

H Enthalpy

Tj Pj

Qj

LDj

VDj

Vj yj

_Lj-1 xj-1

_

Lj xj

_Vj+1 yj+1

_

Fj XF

_


ConvergenceEquations

• Number of equations = NT(2NC+3)

• Example if you have 20 trays and 20 components

• 20(40+3) = 860 equations

• Multiple non-blended petroleum assays feeding a 20-tray column

• 10,000+ equations

• Equations are nonlinear

• All of PRO/II’s distillation methods are iterative

Tj Pj

Qj

LDj

VDj

Vj yj

_Lj-1 xj-1

_

Lj xj

_Vj+1 yj+1

_

Fj XF

_


Non-Convergence

1) Improper Damping

2) Unsuitable Column Algorithm / Estimation Method

3) Initial Estimates are too far / too close to solution

4) Improper or Infeasible Specs

5) Not Enough Iterations

5-Reasons

Damping



ConvergenceNewton’s Method

Solution

f(x)

0

Xx1 x2 xnx0

( )

( )( )

( )( )( )( )0

01

0

01

1

run

rise

xf

xfxx

xf

xfxx

xx

xfxf

xf

nn

nn

nn

nn

n

−=

=−

−=

=

+

+

+


ConvergenceNewton’s Method

“When you throw the football, only three things can happen – and two of em ain’t good”

1) Converge

2) Diverge

3) Oscillate

Woody Hayes, Ohio State Football Coach 1966

The same is true for trying to converge a distillation column


BackgroundPotential Issues

Bad guess converges But better guess fails!

f(x)

x* X

0

Periodic

A B C D


BackgroundDiverging Sequence

xkxk+1 xk+2

F(x)

X


BackgroundDiverging Sequence

Estimates



EstimatesSIMPLE

• Default for columns without a condenser or a reboiler or both• The column profiles are determined by simple material balances.• Temperatures are determined by estimating the product compositions.

CONVENTIONAL• Default for columns with both a condenser and a reboiler• Works well with most columns• Fenske shortcut distillation calculations are used.• Strongly dependent on your product rate estimates

REFINING • Complex refinery columns (e.g., Crude, Vacuum, FCC main fractionator, Coker)• A multi-product shortcut technique developed by Aveva is to generate the

estimates for these columns.

CHEMICAL • Non-ideal thermodynamics (e.g., azeotropic and extractive distillation).• Calculates of a series of adiabatic flashes up and down the column trays

several times as needed to generate a good composition starting profile.• Slow

IEG

Initial Specifications



Feed Location

• Poor location can result in a compositional pinch that cannot be solved

Reflux

• Should be safely above the minimum (simulation can become unstable as the

minimum reflux is approached)

Condenser and Reboiler Issues

• Bubble point condenser with non-condensables

• Fixed duty limitations

• Specifying duties and reflux

SpecificationsInitial Column Setup


Feed Temperature

• Convergence problems in the simulation might represent actual physical problems

Temperature and Product Spec on a Tray

• There is only a narrow range of convergence

Specifying Tray Temperatures

• Use only where composition changes significantly in the column

SpecificationsTemperature


More Than One

• Not all methods will solve (I/O is the best)

• Replace a compositional specification with a RRatio or product rate

Two Compositional Specifications for one Product

• May be impossible

• Has a chance if the relative volatilities are high enough

Same Component Specification in Different Products

• Feasible but difficult

• Has a chance if the relative volatilities are high enough

• Difficult if the concentrations are small

SpecificationsProducts


SpecificationsProducts (cont)

High Purity Specification

• Impurity specs are easier to hit

Compositional and Product Flow

• May clash

• Free one and specify something else

Specifying all Product Flows

• No freedom to establish a material balance

• At least one product flow rate should be allowed to float


SpecificationsProducts (cont)

Infinite Solutions

• Never use 0 or 100 in a specification

Unfeasible

• 300 lbmol/hr propane in the OVHD – only 250 lbmol/hr feed

• Recover specifications are initially safer

Algorithms



AlgorithmsTypes of Distillation

• Simple / Ideal Separation

• Wide and Narrow Boilers

• Azeotropic

• VLE / VLLE

• Electrolytic

• Reactive

• Rate Based

• Liquid-Liquid


AlgorithmsMathematical Models

• Inside-Out (I/O)

• Sure

• Chemdist

• Liquid-Liquid

• Enhanced I/O

• RATEFRAC

• Electrolytic


AlgorithmsMathematical Models

Choosing a column algorithm is based on two factors:

• The thermodynamic complexity of the chemical species

• The mechanical complexity of the column

Identical results are obtained regardless of algorithm used


AlgorithmsInside-Out (I/O)

• Thermodynamically simple but mechanically complex

• Robust

• Fast

• No VLLE

BTMS

LN-2

VN-1

BottomSump Reboiler

Sump

RF

RV

RL

LO

BTMS

LN-2

LN-2

VN-1

VN-1

BottomSump

BottomSump Reboiler

SumpReboilerSump

RF

RF

RV

RV

RL

RL

LO

LO

2

18

20

CRUDE

7

8

11

12

15

16

AGO

1

2

DIESEL

1

2

1

2

1

3

QFZ

RESID

DECANT

AGODRAW

DIESELDRAW

KERODRAW

NAPHTHA

KERORTN

KERODIESELRTN

AGORTN

22

18

20

CRUDE

77

88

1111

1212

15

16

AGO

1

2

11

22

DIESEL

1

2

11

22

1

2

11

22

1

33

QFZQFZ

RESID

DECANT

AGODRAW

DIESELDRAW

KERODRAW

NAPHTHA

KERORTN

KERODIESELRTN

AGORTN



• To converge the inner loop the sum of the errors (Enth+Spec) < 2.0 E-05

• If the inner loop fails to converge and the (Enth+Spec) > 1, look for gross errors

• If the outer loop oscillates - the simple thermo assumption is probably not valid.


AlgorithmsOscillations

Oscillations


AlgorithmsSure Method

• This method uses a modified Newton-Raphson solution technique with equation partitioning (proprietary equation tearing)

• Side columns are solved as recycles

• Somewhat sensitive to initial estimates

• PA’s allowed

• Thermosiphons (as of 9.0)

• VLLE



• Two liquid phases may be considered

• A free water phase and water draws are permitted in columns at trays below the condenser

• Total pumparounds are permitted

• Handles non-ideal interactions

Advantages

Disadvantages

• Slow

• Need accurate estimates

• Side columns are handled as recycle streams


AlgorithmsChemdist

• Mechanically simple – thermodynamically complex

• It uses a modified Newton Raphson algorithm with a proprietary matrix solver developed by SimSci.

• PRO/II checks the VLE trays for possible VLLE behavior. If VLLE behavior is detected, the column may be resolved using VLLE system for these trays.

• No thermosiphons

• No PA’s

• Very sensitive to initial estimates

• Reactive distillation

• Azeotropic distillation


AlgorithmsChemdist

ConvergenceErrors



Convergence ErrorsCheck the Column Profile and Iterations Messages


Convergence ErrorsCheck the Column Profile and Iterations Messages

Partial - Default

All


Convergence Errors

• Unsuitable Column Algorithm

• Initial Estimates are too far / too close to solution

• Improper or Infeasible Specs


Convergence Errors

Not Enough Iterations

• Especially if damping applied

• Check the printout

Bad Actor Components

• Water can seriously upset a VLE method

• Insure that water is not condensing on any trays in your column

PA’s can be Problematic

• Fix rate and duty - calculate return temperature.

• Excess cooling cause drying above PA return

• Remedy: Specify liquid flow above return tray and calculate pumparound duty

Specify Tray

Liquid rate

Declare Duty

as a Variable

Specify Tray

Liquid rate

Declare Duty

as a Variable

Iterations


Convergence Errors

Systems with Highly Non-Ideal Thermodynamics

• You need very good estimates

• Solve with a simple approximation and then use this for a new estimate

Evaluate the Column solution

• What does the temperature/ vapor and liquid flows look like

• What specifications have and have not been met

If the Column is in a Recycle Loop

• Ensure that the vapor and liquid traffic have been established

• Ensure that the recycle loop is at a lower tolerance than the units in the loop

H2


Convergence Errors

Understand how Subsequent runs are Handled in AVEVA PRO/II Simulation

• PRO/II ‘remembers’ product flow rates from previous solutions, and uses those to restart the IEG

• If the previous run did not converge, product flow rates may be very bad, and throw off the IEG for subsequent runs

• Click on ‘Restore Input Data’ in the Input menu item


Convergence Errors

Damping

• Check the messages for a read back on how the algorithm performed

ConvergenceStrategy



Strategy

There are two ways of designing a process model

1) One way is to make it so simple that there are obviously no problems

2) The other way is to make it so complicated that there are no obvious problems

The first is far more difficult


Model Complexity


Start Slowly


Provide Estimates


Strategy

If the column is not solving, provide molar estimates

• Internally distillation algorithms converge on moles because VLE equations are written using moles (not mass or volume)


Strategy

Use Stream / Unit Op Property Tables to facilitate convergence


Strategy

Check the Column Profile


Strategy

Use an Appropriate Algorithm

SWS

Conclusions and Takeaways


• Start off simple

• Ensure you have the correct setup (IEG, Algorithm, Estimates, and Iterations)

• Examine the output for oscillations or divergence – use damping

• Ensure that all trays have vapor and liquid

• 1 800 SimSci1

• [email protected]

This presentation may include predictions, estimates, intentions, beliefs and other statements that are or may be construed as being forward-looking. While these forward-looking statements represent our current judgment on what the future holds, they are subject to risks and uncertainties that could result in actual outcomes differing materially from those projected in these statements. No statement contained herein constitutes a commitment by AVEVA to perform any particular action or to deliver any particular product or product features. Readers are cautioned not to place undue reliance on these forward-looking statements, which reflect our opinions only as of the date of this presentation.

The Company shall not be obliged to disclose any revision to these forward-looking statements to reflect events or circumstances occurring after the date on which they are made or to reflect the occurrence of future events.


Questions?

linkedin.com/company/aveva

@avevagroup

ABOUT AVEVA

AVEVA, a global leader in industrial software, drives digital transformation for industrial organizations managing complex operational processes. Through Performance Intelligence, AVEVA connects the power of information and artificial intelligence (AI) with human insight, to enable faster and more precise decision making, helping industries to boost operational delivery and sustainability. Our cloud-enabled data platform, combined with software that spans design, engineering and operations, asset performance, monitoring and control solutions delivers proven business value and outcomes to over 20,000 customers worldwide, supported by the largest industrial software ecosystem, including 5,500 partners and 5,700 certified developers. AVEVA is headquartered in Cambridge, UK, with over 6,000 employees at 90 locations in more than 40 countries. For more details visit: www.aveva.com


http://www.aveva.com/

AVEVA PI WORLD Distillation Column Convergence in AVEVA ...

Documents

Transcript of AVEVA PI WORLD Distillation Column Convergence in AVEVA ...